Keras.callbacks
my_callbacks = [
tf.keras.callbacks.EarlyStopping(monitor='acc', patience=2),
tf.keras.callbacks.ModelCheckpoint(filepath='model.{epoch:02d}-{val_loss:.2f}.h5',monitor='val_loss',save_best_only=True),
tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss',factor=0.1, patience=10)
]
model.fit(dataset, epochs, batch_size, callbacks=my_callbacks)
- monitor为检测指标,patience即多少epochs未变好就终止训练
- filepath为存储的位置和模型名称,以.h5为后缀,monitor为检测的指标,这里为验证集的成功率,save_best_only代表只保存最优的训练结果。
- factor学习率减小倍数,patience连续批次minitor未下降条件